人工智能培训

搜索

人工智能论文:用于重新识别的时域中的学习特征聚合(Learning Feature Aggregation in Temporal Domain for Re-

[复制链接]
firetea 发表于 2019-3-15 13:24:09 | 显示全部楼层 |阅读模式
firetea 2019-3-15 13:24:09 493 0 显示全部楼层
人工智能论文:用于重新识别的时域中的学习特征聚合(Learning Feature Aggregation in Temporal Domain for Re-Identification)人员重新识别是计算机视觉社区中的标准问题。近年来,车辆重新识别也受到更多关注。在本文中,我们关注这两个任务并提出一种在时域中聚合特征的方法,因为对同一对象进行多次观察是很常见的。聚合基于通过不同权重对特征向量的不同元素进行加权,并且由Siamese网络以端对端方式训练。实验结果表明,我们的方法在车辆和人员重新识别任务上优于其他现有的时域特征聚合方法。此外,为了进一步推动车辆重新识别研究,我们引入了一个新的数据集CarsReId74k。数据集不限于前/后视点。它包含17,681个独特的车辆,73,976个观察轨道和277,236个正面对。 66个摄像机从各个角度捕获数据集。
Person re-identification is a standard and established problem in thecomputer vision community.In recent years, vehicle re-identification is alsogetting more attention.In this paper, we focus on both these tasks and proposea method for aggregation of features in temporal domain as it is common to havemultiple observations of the same object.The aggregation is based on weightingdifferent elements of the feature vectors by different weights and it istrained in an end-to-end manner by a Siamese network.The experimental resultsshow that our method outperforms other existing methods for feature aggregationin temporal domain on both vehicle and person re-identification tasks.Furthermore, to push research in vehicle re-identification further, weintroduce a novel dataset CarsReId74k.The dataset is not limited tofrontal/rear viewpoints.It contains 17,681 unique vehicles, 73,976 observedtracks, and 277,236 positive pairs.The dataset was captured by 66 cameras fromvarious angles.人工智能论文:用于重新识别的时域中的学习特征聚合(Learning Feature Aggregation in Temporal Domain for Re-Identification) MwZBJGwpazgKoPZD.jpg
URL地址:https://arxiv.org/abs/1903.05244     ----pdf下载地址:https://arxiv.org/pdf/1903.05244    ----人工智能论文:用于重新识别的时域中的学习特征聚合(Learning Feature Aggregation in Temporal Domain for Re-Identification)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

firetea当前离线
新手上路

查看:493 | 回复:0

快速回复 返回顶部 返回列表